Работа с пакетами CARET и BORUTA на языке R
2026-05-29
Установка пакетов выполняется один раз.
install.packages(c(
"caret",
"FSelectorRcpp",
"arules",
"Boruta",
"mlbench",
"randomForest",
"rmarkdown",
"knitr"
))library(caret)
library(FSelectorRcpp)
library(arules)
library(Boruta)
library(mlbench)
set.seed(123)
if (!dir.exists("figs")) {
dir.create("figs")
}1. Пакет CARET
names(getModelInfo())## [1] "ada" "AdaBag" "AdaBoost.M1"
## [4] "adaboost" "amdai" "ANFIS"
## [7] "avNNet" "awnb" "awtan"
## [10] "bag" "bagEarth" "bagEarthGCV"
## [13] "bagFDA" "bagFDAGCV" "bam"
## [16] "bartMachine" "bayesglm" "binda"
## [19] "blackboost" "blasso" "blassoAveraged"
## [22] "bridge" "brnn" "BstLm"
## [25] "bstSm" "bstTree" "C5.0"
## [28] "C5.0Cost" "C5.0Rules" "C5.0Tree"
## [31] "cforest" "chaid" "CSimca"
## [34] "ctree" "ctree2" "cubist"
## [37] "dda" "deepboost" "DENFIS"
## [40] "dnn" "dwdLinear" "dwdPoly"
## [43] "dwdRadial" "earth" "elm"
## [46] "enet" "evtree" "extraTrees"
## [49] "fda" "FH.GBML" "FIR.DM"
## [52] "foba" "FRBCS.CHI" "FRBCS.W"
## [55] "FS.HGD" "gam" "gamboost"
## [58] "gamLoess" "gamSpline" "gaussprLinear"
## [61] "gaussprPoly" "gaussprRadial" "gbm_h2o"
## [64] "gbm" "gcvEarth" "GFS.FR.MOGUL"
## [67] "GFS.LT.RS" "GFS.THRIFT" "glm.nb"
## [70] "glm" "glmboost" "glmnet_h2o"
## [73] "glmnet" "glmStepAIC" "gpls"
## [76] "hda" "hdda" "hdrda"
## [79] "HYFIS" "icr" "J48"
## [82] "JRip" "kernelpls" "kknn"
## [85] "knn" "krlsPoly" "krlsRadial"
## [88] "lars" "lars2" "lasso"
## [91] "lda" "lda2" "leapBackward"
## [94] "leapForward" "leapSeq" "Linda"
## [97] "lm" "lmStepAIC" "LMT"
## [100] "loclda" "logicBag" "LogitBoost"
## [103] "logreg" "lssvmLinear" "lssvmPoly"
## [106] "lssvmRadial" "lvq" "M5"
## [109] "M5Rules" "manb" "mda"
## [112] "Mlda" "mlp" "mlpKerasDecay"
## [115] "mlpKerasDecayCost" "mlpKerasDropout" "mlpKerasDropoutCost"
## [118] "mlpML" "mlpSGD" "mlpWeightDecay"
## [121] "mlpWeightDecayML" "monmlp" "msaenet"
## [124] "multinom" "mxnet" "mxnetAdam"
## [127] "naive_bayes" "nb" "nbDiscrete"
## [130] "nbSearch" "neuralnet" "nnet"
## [133] "nnls" "nodeHarvest" "null"
## [136] "OneR" "ordinalNet" "ordinalRF"
## [139] "ORFlog" "ORFpls" "ORFridge"
## [142] "ORFsvm" "ownn" "pam"
## [145] "parRF" "PART" "partDSA"
## [148] "pcaNNet" "pcr" "pda"
## [151] "pda2" "penalized" "PenalizedLDA"
## [154] "plr" "pls" "plsRglm"
## [157] "polr" "ppr" "pre"
## [160] "PRIM" "protoclass" "qda"
## [163] "QdaCov" "qrf" "qrnn"
## [166] "randomGLM" "ranger" "rbf"
## [169] "rbfDDA" "Rborist" "rda"
## [172] "regLogistic" "relaxo" "rf"
## [175] "rFerns" "RFlda" "rfRules"
## [178] "ridge" "rlda" "rlm"
## [181] "rmda" "rocc" "rotationForest"
## [184] "rotationForestCp" "rpart" "rpart1SE"
## [187] "rpart2" "rpartCost" "rpartScore"
## [190] "rqlasso" "rqnc" "RRF"
## [193] "RRFglobal" "rrlda" "RSimca"
## [196] "rvmLinear" "rvmPoly" "rvmRadial"
## [199] "SBC" "sda" "sdwd"
## [202] "simpls" "SLAVE" "slda"
## [205] "smda" "snn" "sparseLDA"
## [208] "spikeslab" "spls" "stepLDA"
## [211] "stepQDA" "superpc" "svmBoundrangeString"
## [214] "svmExpoString" "svmLinear" "svmLinear2"
## [217] "svmLinear3" "svmLinearWeights" "svmLinearWeights2"
## [220] "svmPoly" "svmRadial" "svmRadialCost"
## [223] "svmRadialSigma" "svmRadialWeights" "svmSpectrumString"
## [226] "tan" "tanSearch" "treebag"
## [229] "vbmpRadial" "vglmAdjCat" "vglmContRatio"
## [232] "vglmCumulative" "widekernelpls" "WM"
## [235] "wsrf" "xgbDART" "xgbLinear"
## [238] "xgbTree" "xyf"x <- matrix(rnorm(50 * 5), ncol = 5)
colnames(x) <- paste0("X", 1:5)
y <- factor(rep(c("A", "B"), 25))jpeg("figs/caret_pairs.jpg", width = 1200, height = 900, quality = 95)
print(featurePlot(x = x, y = y, plot = "pairs", auto.key = list(columns = 2)))
dev.off()## png
## 2jpeg("figs/caret_density.jpg", width = 1200, height = 900, quality = 95)
print(featurePlot(x = x, y = y, plot = "density",
scales = list(x = list(relation = "free"))))
dev.off()## png
## 2jpeg("figs/caret_box.jpg", width = 1200, height = 900, quality = 95)
print(featurePlot(x = x, y = y, plot = "box"))
dev.off()## png
## 2
Вывод: признаки сгенерированы случайно, поэтому
классы A и B на графиках сильно перекрываются.
Явного разделения классов по признакам не наблюдается.
2. Важность признаков для iris
Обычный пакет FSelector на данном компьютере не
запускается из-за ошибки Java, поэтому использован совместимый пакет
FSelectorRcpp.
data(iris)ig <- information_gain(Species ~ ., data = iris, type = "infogain")
gr <- information_gain(Species ~ ., data = iris, type = "gainratio")
su <- information_gain(Species ~ ., data = iris, type = "symuncert")
rel <- relief(Species ~ ., data = iris, neighboursCount = 5, sampleSize = 20)
importance <- data.frame(
Feature = ig$attributes,
Information_Gain = ig$importance,
Gain_Ratio = gr$importance,
Symmetrical_Uncertainty = su$importance,
Relief = rel$importance
)
importance[order(-importance$Information_Gain), ]## Feature Information_Gain Gain_Ratio Symmetrical_Uncertainty Relief
## 4 Petal.Width 0.9554360 0.8713692 0.8705214 0.3502083
## 3 Petal.Length 0.9402853 0.8584937 0.8571872 0.3440678
## 1 Sepal.Length 0.4521286 0.4196464 0.4155563 0.1645833
## 2 Sepal.Width 0.2672750 0.2472972 0.2452743 0.1356250barplot(
importance$Information_Gain,
names.arg = importance$Feature,
las = 2,
col = "lightblue",
main = "Важность признаков iris",
ylab = "Information Gain"
)
Вывод: наиболее важными признаками для классификации
ирисов являются Petal.Length и Petal.Width.
Признаки Sepal.Length и Sepal.Width менее
информативны.
3. Дискретизация признаков
sl <- iris$Sepal.Lengthd_interval <- discretize(sl, method = "interval", breaks = 3)
table(d_interval)## d_interval
## [4.3,5.5) [5.5,6.7) [6.7,7.9]
## 52 70 28d_frequency <- discretize(sl, method = "frequency", breaks = 3)
table(d_frequency)## d_frequency
## [4.3,5.4) [5.4,6.3) [6.3,7.9]
## 46 53 51d_cluster <- discretize(sl, method = "cluster", breaks = 3)
table(d_cluster)## d_cluster
## [4.3,5.33) [5.33,6.27) [6.27,7.9]
## 46 53 51d_fixed <- discretize(sl, method = "fixed", breaks = c(5.5, 6.5))
table(d_fixed)## d_fixed
## [5.5,6.5]
## 68head(data.frame(
Sepal.Length = sl,
interval = d_interval,
frequency = d_frequency,
cluster = d_cluster,
fixed = d_fixed
))## Sepal.Length interval frequency cluster fixed
## 1 5.1 [4.3,5.5) [4.3,5.4) [4.3,5.33) <NA>
## 2 4.9 [4.3,5.5) [4.3,5.4) [4.3,5.33) <NA>
## 3 4.7 [4.3,5.5) [4.3,5.4) [4.3,5.33) <NA>
## 4 4.6 [4.3,5.5) [4.3,5.4) [4.3,5.33) <NA>
## 5 5.0 [4.3,5.5) [4.3,5.4) [4.3,5.33) <NA>
## 6 5.4 [4.3,5.5) [5.4,6.3) [5.33,6.27) <NA>Вывод: метод interval делит значения на
интервалы равной ширины, frequency делает группы примерно
равными по числу объектов, cluster использует
кластеризацию, а fixed применяет заранее заданные
границы.
4. Выбор признаков с помощью Boruta
data("Ozone", package = "mlbench")
ozo <- na.omit(Ozone)boruta_ozo <- Boruta(V4 ~ ., data = ozo, doTrace = 0, maxRuns = 100)
boruta_final <- TentativeRoughFix(boruta_ozo)## Warning in TentativeRoughFix(boruta_ozo): There are no Tentative attributes!
## Returning original object.boruta_final## Boruta performed 21 iterations in 0.1828592 secs.
## 9 attributes confirmed important: V1, V10, V11, V12, V13 and 4 more;
## 3 attributes confirmed unimportant: V2, V3, V6;attStats(boruta_final)## meanImp medianImp minImp maxImp normHits decision
## V1 0.40539847 0.40642988 0.33443318 0.46794786 1.00000000 Confirmed
## V2 0.04694280 0.03996016 -0.01933979 0.11724390 0.04761905 Rejected
## V3 -0.05868909 -0.07569424 -0.11701706 0.03393198 0.00000000 Rejected
## V5 0.38618563 0.39003277 0.34188322 0.42571719 1.00000000 Confirmed
## V6 0.04184227 0.03949771 -0.02048419 0.13561582 0.14285714 Rejected
## V7 0.51827308 0.51510008 0.44684417 0.59395833 1.00000000 Confirmed
## V8 0.77327105 0.78086300 0.69631396 0.82506181 1.00000000 Confirmed
## V9 0.87164021 0.86647893 0.79318976 0.93859006 1.00000000 Confirmed
## V10 0.44857358 0.44882788 0.41852372 0.48910026 1.00000000 Confirmed
## V11 0.54561567 0.55227270 0.44966453 0.63220513 1.00000000 Confirmed
## V12 0.65892780 0.64623327 0.58965492 0.71349070 1.00000000 Confirmed
## V13 0.40205675 0.40106143 0.35016242 0.45802074 1.00000000 Confirmedjpeg("figs/boruta_ozone_boxplot.jpg", width = 1200, height = 900, quality = 95)
plot(boruta_ozo, las = 2, cex.axis = 0.7)
dev.off()## png
## 2
getSelectedAttributes(boruta_final, withTentative = FALSE)## [1] "V1" "V5" "V7" "V8" "V9" "V10" "V11" "V12" "V13"Вывод: признаки со статусом Confirmed
являются значимыми для предсказания переменной V4. Признаки
со статусом Rejected можно исключить из модели.